Thanks Vinod. Sure will do. May I ask what's different with this?
-Nalini From: vi...@twitter.com [mailto:vi...@twitter.com] On Behalf Of Vinod Kone Sent: Sunday, August 11, 2013 8:35 PM To: user@mesos.apache.org Subject: Re: Mesos slave not starting up Can you try our new instructions at https://github.com/mesos/hadoop ? On Sun, Aug 11, 2013 at 7:19 PM, Johnas, Nalini <njoh...@ebay.com<mailto:njoh...@ebay.com>> wrote: Hi Vinod, I tried everything suggested, still running into the same problem with TASK LOST and there is no executor logs created. One quick question, Is there any restriction with the Hadoop location , does it need to be under mesos build directory. Here's where I have these installed. Mesos build is under : /home/njohnas_dev/mesos-testing/build Hadoop home is under : /home/njohnas_dev/mesos-testing/hadoop Also I don't mind driving up to a Starbucks closer to you, if you can spare like 1/2 hour with your busy schedule to go over my setup and help resolve this issue. (or) open to other suggestions as well. Let me know. Thanks Nalini From: Johnas, Nalini [mailto:njoh...@ebay.com<mailto:njoh...@ebay.com>] Sent: Tuesday, August 06, 2013 11:59 PM To: <user@mesos.apache.org<mailto:user@mesos.apache.org>> Cc: user@mesos.apache.org<mailto:user@mesos.apache.org> Subject: Re: Mesos slave not starting up Thanks Vinod that's helpful. I suspect it could be the hadoop path. Let me give this a try. Nalini Sent from my iPad On Aug 6, 2013, at 11:52 PM, "Vinod Kone" <vinodk...@gmail.com<mailto:vinodk...@gmail.com>> wrote: An executor terminated as soon as it's launched is indicative of slave being unable to fetch/launch the executor. In the case of hadoop framework, If your executor sandbox doesn't have a hadoop.tar.gz or hadoop directory, that means the slave is unable to fetch the executor. It likely means the hdfs url for the executor specified in mapred-site.xml is wrong or inaccessible to the slave. Also ensure that 'hadoop' command is in the PATH of the slave (or specified via --hadoop_home slave flag), because the slave fetches hadoop executor by simply doing 'hadoop fs -copyToLocal <executor uri> <executor sandbox>'. HTH, On Tue, Aug 6, 2013 at 11:14 PM, Johnas, Nalini <njoh...@ebay.com<mailto:njoh...@ebay.com>> wrote: Hi Vinod, Yes. Exactly it gets lost as soon as it is lost and I am trying to figure out why? There are no logs in the executor which makes it difficult to debug. What are the potential root causes that could yield to the task getting lost as soon as it is launched? I could deep dive in that direction. -Nalini From: vi...@twitter.com<mailto:vi...@twitter.com> [mailto:vi...@twitter.com<mailto:vi...@twitter.com>] On Behalf Of Vinod Kone Sent: Sunday, August 04, 2013 10:48 PM To: user@mesos.apache.org<mailto:user@mesos.apache.org> Subject: Re: FW: Mesos slave not starting up Was the syslog in one of the executor sandboxes? From the slave log you showed here, it looked like the executor went LOST as soon as it was launched (i.e., it never registered with the slave) but the syslog shows the executor came up? The executor sandbox in this case would be /tmp/mesos/slaves/201308040150-3892119818-5051-11035-0/frameworks/201308040150-3892119818-5051-11035-0000/executors/executor_Task_Tracker_115/runs/d9094b15-540e-4370-a5b5-042b8c5ae6fa